-
DOE/MCT%074-95/CO49 1
Fundamental Studies for Sol-Gel Derived Gas-Separation
Membranes
Authors:
C. Jeffiey Brinker R. Sehgal
Contractor:
Sandia National Laboratory P.O. Box 5800 Albuquerque, New Mexico
87185-5800
Contract Number:
ALw-938 L 8 5 0 0 0
Conference Title:
Natural Gas RD&D Contractor's Review Meeting
Conference Location:
Baton Rouge, Louisiana
Conference Dates:
April 4 - 6, 1995
Conference Sponsor:
Co-Hosted by Department of Energy (DOE) Morgantown Energy
Technology Center Morgantown, West Virginia and Southern University
and Agricultural and Mechanical College Baton Rouge, Louisiana
-
DISCLAIMER
Portions of this document may be illegible in electronic image
products. Smages are produced from the best available original
document.
-
DISCLAIMER
This report was prepared as an account of work sponsored by an
agency of the United States Government. Neither the United States
Government nor any agency thereof, nor any of their employees,
makes any warranty, express or implied, or assumes any legal
liability or responsibility for the accuracy, completeness, or
usefulness of any information, apparatus, product, or process
disclosed, or represents that its use would not infringe privately
owned rights. Reference herein to any specific commercial product,
process, or service by trade name, trademark, manufacturer, or
otherwise does not necessarily constitute or imply its endorsement,
recommendation, or favoring by the United States Government or any
agency thereof. The views and opinions of authors expressed herein
do not necessarily state or reflect those of the United States
Government or any agency thereof.
This report has been reproduced directly from the best available
copy.
Available to DOE and DOE contractors from the Office of
Scientific and Technical Information, 175 Oak Ridge Turnpike, Oak
Ridge, TN 37831; prices available at (6 15) 576-840 1.
Available to the public from the National Technical Information
Service, U.S. Department of Commerce, 5285 Port Royal Road,
Springfield, VA 22 16 1 ; phone orders accepted at (703)
487-4650.
-
P l Fundamental Studies for Sol-Gel Derived Gas-Separation
Membranes
CONTRACT INFORMATION
Contract Number
Contractor Sandia National Laboratories P.O. Box 5800
Albuquerque, NM 87 185-5800 (505) 272-7627 (telephone) (505)
272-7304 (telefax)
Harold D. Shoemaker Contractor Project Manager
Principle Investigators C. Jeffrey Brinker R. Sehgal
venkat K. Venkataraman Margaret Kotzalas
July 1, 1994 to Sept. 30, 1995
METC Project Manager
Period of Performance
Schedules and Milestones FY 95 Program Schedule
........................................................................
S O N D J F M A M J J A S
pore size reduction of microporoussilica membranes
deposition of titania modified silica membranes
single and two gas transport measurements - of silica and
titania modified silica membranes as a function of temperature
Stability analysis of the membranes
-----_---____-______----------------------------------------------------
ABSTRACT observed that both partial sptering of membranes and
development of larger capillary stresses during membrane formation
lead to consolidation of the membrane structure as evidenced by
increased ideal separation factors, e.g. W O ~ / C ~ > 250 over
the temperature range of 160 to 220 "C. Surface derivatization was
also shown to be an effective
We prepared silica membranes using sol-gel techniques and
explored the effects of post- deposition sintering, CaPfilsY
Stresses developed during drying, and surface derivatization of the
membranes with titanium iso-propoxide. We
/
2
-
tool to reduce the membrane pore size in an angstrom by angstrom
fashion, resulting in comparable separation factors. What’s more,
the altered pore surface chemistry of Ti02 derivatized membranes
may lead to improved stability and impart catalytic properties to
the membrane surface.
OBJECTIVES
The primary objective of this study is to prepare supported
inorganic membranes suitable for gas separatiodgas purification
applications in the processing of natural gas. The design criteria
for such membranes include: high selectivity and high flux combined
with excellent chemical, thermal and mechanical stability.
BACKGROUND INFORMATION
In the processing of natural gas, membranes have been proposed
as replacements for more energy intensive purification processes
for the removal of H20, C02, H2S, and higher hydrocarbons from
methane. Due to their inherent thermal, mechanical, and chemical
stability, inorganic membranes are attractive for such
applications. For practical gas separations, an inorganic membrane
must combine high flux J with a high selectivity factor a. The
membrane flux for Knudsen transport is proportional to the pore
radius rp divided by the film thickness h:
where E is the porosity, AP is the pressure drop across the
membrane, and z* is the tortuosity. Membrane selectivity is related
to pore size and pore size distribution. When the pore size is
reduced below the mean free path of a gas molecule, gas transport
occurs by Knudsen diffusion, and the separation factor for a binary
gas mixture depends on the inverse ratio of the respective
molecular weights. For example, the Knudsen selectivity factor a K
for a Hem2 mixture is 2.65. Much greater separation factors
(approaching -) are achieved when the pore size is
reduced sufficiently and the pore size distribution is
sufficiently narrow to admit and exclude molecules on the basis of
size, a mechanism referred to as molecular sieving. Thus to combine
high flux with high selectivity it is necessary to prepare
extremely thin, porous films with a narrow distribution of
extremely small pores (rp - molecule size) and no cracks or other
imperfections that would serve as large pores and diminish
selectivity.
The traditional approach to the preparation of thin inorganic
membranes with controlled pore sizes is to deposit (slip-cast
and/or dip-coat) particulate sols with narrow particle size
distributions on porous supports.lT23 If aggregation is avoided,
the pore size of the membrane is controlled by the particle size of
the sol - smaller particles yield smaller pores. An advantage of
this approach is that the porosity of the membrane, which dictates
its flux (Es. 1) is independent of the particle size. So-called
Knudsen membranes prepared from particulate sols have been
successfully demonstrated by many research groups and are now
commercially available.4 However due to problems with cracking5 and
pore size.coarsening at elevated temperatures,6 it has proven
difficult to prepare molecular sieving membranes using this
traditional approach.
This paper considers the use of polymeric silicate sols to
prepare thin supported membranes, where the distinction between
particles and polymers, made on the basis of NMR and SAXS
measurements,6 is that the polymeric sols contain no regions of
fully condensed silica. The idea of this alternative approach is to
utilize the fractal properties of polymeric sols to favor polymer
interpenetration during membrane deposition. This concept is
qualitatively understood on the basis of the following relationship
describing the mean number of intersections MI ,2 of two mass
fractal objects of size R placed in the same region of space:7
where D1 and D2 are the r@spective mass fractal dimensions and d
is the dimension of space, 3. From Equation 2, we see that if D1 +
D2 < 3, the probability of intersection decreases indefinitely
with R. This suggests that as D is reduced polymers should more
easily interpenetrate as they are concentrated during
slip-castingdip-coating. If
-
compliant, such interpenetrating networks should collapse during
drying to create pores of molecular dimensions appropriate for
molecular sieving.
As reported in last year's Natural Gas R & D Contractors
Review Meeting report, a practical advantage of this polymer
approach with respect to the preparation of supported membranes is
that the final pore size should be independent of the polymer size.
Thus aging can be used to grow the polymers large enough to be
trapped on the support surface during membrane deposition (avoiding
the creation of thick often imperfect membranes by so- called "pore
plugging" (see Figure l)) without detrimentally increasing the pore
size. Also a broadening of the polymer molecular weight
distribution should not necessarily be manifested as a broadening
of the pore size distribution.
40Aoore3-4umCaver
/
RGccr , , Constrained& mkap
eff ectivememb m e Fore plugging. Thidter
By%+.
Rc -rp Somepenetration
depen dingon condensationrate
Figure 1. Schematic illustration of the effect of the relative
sizes of polymer and
support pores on membrane formation
A second advantage of this approach is that polymeric sols
generally lead to the formation of amorphous membranes that do not
exhibit grain growth or phase transformations during heating (as is
often observed for particulate sols). Thus compared to particulate
membranes, the pore dimensions of polymeric membranes should
exhibit improved thermal stability.
A possible disadvantage of this approach is that small pore
sizes are achieved at the expense of pore volume. This limitation
may be overcome by the
use of organic molecules or ligands as pore templates.
Last year we reported on the effects of sol- aging conditions
and condensation rates on the microstructure and transport behavior
of supported silica membranes and demonstrated a template strategy
as a means of independently controlling pore size and pore volume.
This year we report on the effects of post-deposition sintering,
capillary stress during membrane deposition and a derivatization
technique designed to develop truly molecular sieving
membranes.
Partial Sintering
Partial sintering of an amorphous silica membrane is expected to
further collapse the structure at higher sintering temperatures,
leading to a narrower pore size distribution and/or smaller pore
sizes. The collapse of the structure is accompanied by a large
reduction in porosity and a loss of hydroxyl coverage on the silica
structure.6 The loss of hydroxyl coverage is also expected to make
the structure slightly hydrophobic and thus more stable to attack
by a humid environment.
Drying Stress
w Adsorption A Desorption
5
w Adsorption 4 A Desorption
'h 3 ; a m 2 :
1 - z A E = * ,
A & A 0 , . . . I . . I . . , . . . ,*:. . 0 0.2 0.4
PIP 0 0.6 0.8 1
Ir Figure 2. Stress generated in an 'as
deposited' A2 film vs relative pressure
Recently Josh Samuel* quantified the capillary stress
development in microporous sol-gel derived
-
silica films. The results shown for an 'as deposited' A2 film in
Figure 2 indicate that very large capillary stresses (of the order
5 kbar) are developed at low relative pressures of water in the
overlying gas. These pressures are exerted on the solid phase
during film deposition causing collapse of the gel network and as
we report in the Results section a reduction in pore size and/or
narrowing of the pore size distribution leading to enhanced
separation factors.
Surface Derivatization
Reaction of a hydroxylated silica surface with a dilute solution
of Ti (0 i-Pr)4 can lead to two possible reactions: 1) alcohol
condensation between the surface hydroxyl and the alkoxy group
(Equation 3) and 2) alcoholysis of the siloxane (Si- 0-Si) bonds
(Equation 4).
&-OH + i-Pr-0-Ti= + ESi-O-TiE + i-Pr-OH (3) Si-0-Si= +
i-&-OH + Si -OH + i-Pr 0-Si= (4)
Titanium &oxides are sufficiently electropositive to undergo
reaction 3 even in the absence of a catalyst. Furthermore, the
alcoholysis reaction is relatively slow.6 As a result, the reaction
with Ti(0 i-Pr)4 will presumably lead to the deposition of a
Si-0-Ti layer on the silica surface according to Reaction 3.9 Apart
from the above two reactions some isopropoxide may be adsorbed on
the hydrogen bonded surface hydroxyls associated with the silica
membranes. These adsorbed molecules can subsequently condense
during the calcination stage to form Si-0-Ti bonds similar to
Reaction 3 with the hydrogen bonded intermediate step.
-According to Srinivasan et. al.9 silica surfaces with a larger
fraction of hydrogen bonded hydroxyls have a higher loading of
titania due to the condensation of adsorbed precursor. The hydrogen
bonded surface hydroxyls were found to be the preferred adsorption
sites for the precursor as per infrared spectroscopy.
This titania layer has the potential to improve the separation
properties and the hydrothermal stability of the membrane.1° The
titania membrane also imparts catalytic properties to the membrane,
or alternatively, this membrane can be used as a
high surface area support for other catalysts such as vanadia.9
These modified membranes are very good candidates for membrane
reactor applications.
PROJECT DESCRIPTION/EXPERIMENTAL DETAILS
Preparation of Coating Solutions
The silica sols were prepared in a two step process referred to
as A2**.11 In the first step an A2** stock solution was prepared
with TEOS:EtOH:H20:HCl ratio of 1:3.8:1.1:5 X 10-5. and refluxed
for 90 min. at 60 "C. Prior to the second step of sol formation,
the stock solution was maintained in a freezer at -30 "C. In the
second step additional water and acid was added at room
temperature, resulting in final molar ratios of TEOS:EtOH:H20:HCI
of 1.0:3.8:5.1:0.004. The sol was shaken for 15 minutes using a
wrist action shaker and aged in a 50 "C oven to determine the gel
time. The coating sol was prepared by diluting the A2** sol aged
for t/t el = 0.24 (t el = 99 hours), with twice the vo%me of 208
proof ethanol soon after removal from the oven. The coating sols
are stable for over 2 months when stored at -30 "C.
The titania modification solution for the silica membranes was
prepared by mixing 5 vol. % Ti (0 i-Pr)4 in tetrahydrofuran (THF).
Hexane and toluene are used for removal of excess precursor after
the titania modification step. The Ti (0 i-Pr)4, THF, hexane and
toluene must be freshly distilled to remove any trace amounts of
water as the precursor is highly susceptible to hydrolysis.
Membrane Deposition
Membranes were deposited on commercially available Membralox@
supports (supplied by U. S. Filter in 25 cm lengths). Supports were
cut into 5 cm sections, using a diamond wafering saw. Each uncoated
support was pre-talcined to 400 "C or 550 "C depending on the final
membrane calcination temperature.
Before coating, the supports were cleaned with CO;! snow using a
SNOGUNTM cleaner. The silica membranes were deposited using a
Compumotorm
-
linear translation stage in a glove box, which was continuously
backfilled with dry N2 from a liquid N2 dewar to maintain a clean
and dry environment. Purging time of the dry box was used to
control the relative pressure of water. The support was dip- coated
at a constant immersion and removal rate of 20 crn/min. The support
was held immersed in the sol for 100 seconds prior to removal.
After dip- coating the membrane was allowed to dry for 15 minutes
in the glove box before calcination in air to 400 OC or 550 "C for
3 hours with a heating and cooling rate of 1 "C/min. After
calcination the membranes were always stored in air at 150 "C.
membranes calcined to 550 OC. For the modification experiments
the dry box was maintained at < 10 ppm of water. The water
content in the box is checked with a Tic4 solution by opening the
bottle periodically while the box is being purged with dry
nitrogen. The chamber has less than 10 ppm water when the Tic4
solution no longer fumes when opened. At this stage, the container
of the modification solution is opened and the membrane is lowered
into the solution. The membrane is allowed to react with the
solution for five minutes and then withdrawn. The reaction is
followed by repeated hexane and toluene washing.
The membrane was subsequently removed from the dry box and
calcined at 400 "C for 3 hours using heating and cooling rates of 1
"Umin.
Titania modification was performed on A2**
Characterization
Membranes. The membranes were characterized using a manual or an
automated single gas permeability measurement system. The automated
system is capable of measuring the permeance of six different gases
in succession at various temperatures and pressures. The system is
set-up to measure the membrane transport properties over the
temperature range 25-220 "C and pressure range of 12 to 92 psia.
The series of gases used in the permeability experiments was He,
H2, N2, C3H6, C02, CH4, and SF6. In case the permeance was below
the accurate detection range of the mass flow controllers (MFCs)
connected to the system, the flow data was collected by running the
system in a semi- automated mode in which gas flux was measured
manually with bubble meters.
Adsorption. To obtain the adsorption measurements on bulk silica
and modified samples, freshly gelled silica sols were dried in an
explosion proof oven at 50 "C. The gel was allowed to dry for
approximately 1 day and subsequently calcined to 400 or 550 "C for
3 hours with a heating and cooling rate of 1 "C/min. The calcined
xerogels thus obtained were used for N2 adsorption measurements at
77 K. The bulk gels were modified with titania following the same
procedure outlined earlier.
The adsorption data was collected over the relative pressure
range of 10-6 to 1 using a Micromeritics ASAP 2000.
Ellipsometry. Companion coatings to supported membranes were
prepared on dense single crystal silicon wafers for ellipsometry
analysis. A GaertnerB model L116C ellipsometer, with a He-Ne laser
light source, was used to derive the thickness and refractive
indices of the silica films. The film porosity was calculated using
the Lorentz-Lorenz model5 (equation 5) assuming a skeletal
refractive index of 1.46 for silica,
where nf is the film refractive index, Vs is the volume fraction
solids, and n, is the refractive index of the solid skeleton.
RESULTS
Effect of sintering
Sintering was studied as means of reducing pore size and
narrowing pore size distribution for bulk, thin film and membrane
specimens. Figure 3 plots the adsorption isotherms for bulk A2**
xerogels calcined to two Vferent temperatures under similar
conditions. The isotherms clearly indicate a sharp decrease in the
surface area and volume % porosity for the 550 "C sample compared
to the 400 "C sample. The sharper isotherm for the A2** sample
calcined to 550 OC also indicates the loss of the larger pores,
-
consistent with a narrower pore size distribution. Similar
results were observed in case of thin A2** films deposited on dense
silicon wafers as summarized in Table 1. We see that the film
thickness as well as the volume % porosity is reduced with
increasing calcination temperature indicative of a more
consolidated structure.
140 * -120
100
80 2 9 60
40
e4 . U
- 2 P 20
0
+A2" Gel, 550 oc +AZ** TO, Modif. .
BET Surracc vdume R, Ares (m'lg) PaMitY
A2" 400 "C 459 n AZ** 550 *C Ti Modif 20 is A2** 550 "C 2s
17
A A A - A T . _ - . . -_ T . T
0 0.2 0.4 0.6 0.8 1
p/P e
Figure 3. Surface area and porosity comparison, pure silica A2**
and
modified A2** Gels
Table 1. Film Thickness and Porosity comparison of single coated
A2** films
Sintered at Different Temperatures
1.375 1300 1.410
Table 2. Permeability Results on A2** Membranes Sintered at
Different
Temperatures
1 I H e I He/ I He/ I He/ I He/ I COd(I Perm. N2 C q CH4 sF6
aH';
Single layer 0.0085 2.0 2.4 1.6 4.0 0.66
Double layer 0.0021 6.9 0.74 2.6 11.8 3.5
Ideal Knudsen - 2.65 3.31 2.0 6.0 0.6 membrane, calc.
400 "C
Single or double layer membranes were prepared using the A2**
sol composition and calcined to 400 O C or 550 "C. Table 2
summarizes the permeability results obtained at 25 "C. We can
clearly see that the second A2** coating calcined to 400 "C leads
to a four fold loss in He permeance accompanied by an increase in
gas selectivities above the ideal Knudsen values for the various
gas pairs. The second coating presumably acts as a healing layer
for the first membrane coating and leads to blockage of the defects
or larger pores in the fist coating. Instead if a single coating of
the A2** sol is calcined to 550 "C we obtain ideal separation
factors larger than for the double-layer A2** membrane calcined to
400 O C along with higher permeance.
These observations lead to the following conclusions: 1) Partial
sintering is effective in reducing pore size and/or narrowing PSD
resulting in greater selectivity and lower flux. 2) Partial
sintering enables the achievement of molecular sieving using a
single membrane layer, obviating the need of a second layer and
resulting in greater flux than observed for double layer
membranes.
Effect of Drying Chamber Humidity
AE (Bcal/mol) a,,w@ 200°C +He 3.7 *% 2.8
0.0015 , . . . . , . . . . , . . . . I . . . . , . . . . , h
8 E
0 50 100 150 200 250 Temperature ("C)
Figure 4. Permeability vs temperature, A2** membrane dyosited in
drying
chamber with > 10 ppm water
Figure 4 plots the typical permeance of the various gases
through an A2** membrane calcined to 550 OC versus measurement
temperature over the
-
temperature range, 30 to 220 "C. This membrane was prepared in
the dry box with a short purge time causing > 10 ppm of water in
the box. The average pressure across the membrane for all the data
points is 52 psia with a pressure drop of 80 psi. Ideal separation
factors determined at 200 "C are indicated on the plot along with
the apparent activation energies for transport. The large
activation energies for He and H2 indicate a very narrow pore size
distribution for the membrane. The high separation factors are also
an indication of a narrow PSD and suggest that the membrane has a
very small fraction of pores larger than the size of a methane
molecule. If we can further reduce the pore size and narrow the
pore size distribution we expect to achieve membranes that exhibit
perfect molecular sieving behavior.
We have discussed earlier that, one of the ways to collapse the
silica structure, is by increasing the capillary stress at the time
of membrane formation. We have also discussed in Figure 2 that the
stress can be increased by reducing the drying chamber
humidity.
+He -H2
0 50 100 1.50 200 250 Temperature ("C)
Figure 5. Permeability vs temperature, A2** membrane deposited
in drying
chamber with < 10 ppm water
Figure 5 and 6 plot the permeance and selectivity results for an
A2** membrane prepared when the water content in the coating
chamber was reduced below 10 ppm before membrane deposition. The
two plots indicate some very significant results. We see that there
is no CHq flux through the membrane over the complete
temperature range. There is no N2 flux through the membrane
below 180 "C and no C02 flux below 80 "C. The flow of these gases
was much below the minimum detection limit of the instruments (<
- 10-7 cm3/cm2-s-cm Hg). Thus we can easily separate various sets
of gases by operating the membrane at selected temperatures and
achieve almost perfect selectivity. An order of magnitude
I A HdCO 0 He/N
HZrn 2
A A ,
A A
1
0 50 100 150 200 250 Temperature ("C)
Figure 6. Ideal separation factor vs temperature, A2** membrane
deposited in
drying chamber with e 10 ppm water
increase in N2 permeance between 160 and 190 "C indicates that
the activation energy for transport is very large over this
temperature range. A similar result is seen for the case of H2 over
the temperature range, 40 to 80 "C.
membranes in a one step process which have true molecular
sieving properties.
membrane at 40 "C (2 x 10-5 cm3/cm2-s-cm Hg) with the baseline
He flux through vitreous silica at 27 "C (7 x 10-9 cm3/cm2-s-cm Hg)
indicates that the flux through our membrane is about 4 orders of
magnitude greater than the flux through dense silica. This
comparison indicates that the fluxes that we are measuring for our
membranes must be due to the presence of well defined
molecular-sized pores.
We have therefore been able to prepare
A comparison of He flux through this
7
-
Titania Modified Silica Membranes
The comparison of the N2 adsorption isotherms of an A2** bulk
gel and an A2** bulk gel modified with Ti@ in Figure 3 indicates
that the Ti02 modification step causes a greater reduction in
porosity than partial sintering.
Figure 7 shows permeance versus temperature for an A2** membrane
deposited at >10 ppm water and modified twice with Ti03 For this
membrane we have completely excluded CHq and N2 over the
temperature range 30 - 220 "C. C02 flux is not detectable below 180
"C, and H2 flux is not detectable below 160 "C.
10-3 ~~"~~~~~~
+He -H,
t C O ,
1 0 - 7 " " ~ ' " ' . ' " ' " ~ " " ~ " ' I 0 50 100 150 200
250
Temperature ("C)
Figure 7. Permeability vs temperature, titania modified A2**
membrane, with two
T i 0 2 modifications
A closer comparison of this membrane with the membrane in Figure
5 indicates that He permanence through the two membranes over the
complete temperature range are comparable. The difference is in the
flux of the other gases. The titania modified membrane seems to
have a smaller number of large pores (we observe a larger N2
permeance through the pure silica membrane). The activation
energies of the gases through these membranes are similar to the
activation energies for most zeolites and carbon molecular sieves
with pore diameters of the order of 4 to 5 A.
SUMMARY/FUTURE WORK
Improvements to the membrane deposition process have led to
significant increases in membrane performance this year. Through
reduction of the relative pressure of water used in the membrane
deposition and drying ambient, we have increased the drying stress
exerted on the gel network leading to a reduction in pore size
and/or a narrowing of the pore size distribution of the resulting
membranes. This is evident from the improvements noted in the
performance of single- - membranes processed to 550 "C (O&/N~
> 250 below 160 "C compared to 9.0 (last year) and %02/cm >
250 over the temperature range 80- 220 "C compared to
(&-02/crra = 1.54 (last year) with He permeance 2 x
cm3/cm2-s-cm Hg at 40 "C). We have also demonstrated a surface
derivatization technique whereby a hydroxylated silica membrane is
reacted with a monomeric titanium alkoxide to reduce the pore size
and modify the pore surface chemistry of a thin region of the
membrane surface. This approach has resulted in %02/cm > 250
above 160 "C with He permeance = 3 x 10-5 cm3/cm2-s-cm Hg at 40
"C.
The stability of the membranes presented in this paper are
currently being studied. The titania modified membranes seem to be
a better candidate for membrane application due to the known
instability of sol-gel derived ultramicroporous silica
structures.
In future we will continue working with the strategies presented
in this paper to further optimize the conditions for membrane
deposition and improve the permeance of the faster permeating
gases.
ACKNOWLEDGEMENTS
Portions of this work were supported by the National Science
Foundation, the Electric Power Research Institute, the Gas Research
Institute, Morgantown Energy Technology Center, and the Department
of Energy Basfc Energy Sciences Program. We are grateful to Jan
Ellison for building the automated gas permeability system. Sandia
National Laboratories is a U. S. Department of Energy facility
supported by Contract Number DE-AC04-76-DP00789.
-
REFERENCES
A.F.M. Leenaars, K. Keizer and A. J. Burggraaf, J. Mat. Sci.,
10, 1077-1088 ( 1 9 84). A. Larbot, J.A. Alary, J.P. Babre, C.
Guizard and L. Cot, "Better Ceramics Through Chemistry 11", Eds.
C.J. Brinker, D.E. Clark and D.R. Ulrich, Mat. Res. SOC. Symp.
Proc., Vol. 73, Mat. Res. SOC., Pittsburgh, PA (1986), 659. M.A.
Anderson, M.J. Gieselman and Q. Xu, J. Membrane Sci., 39,243-258
(1988). J. Guibaud in ICIM-89, "The First International Conference
on Inorganic Membranes, eds. L. Cot and J. Charpin, Montpellier,
France, 1989. T.J. Garino, Ph.D. Dissertation, MIT., Cambridge, MA,
1986. C.J. Brinker and G.W. Scherer, "Sol-Gel Science", Academic
Press, San Diego (1990). B.B. Mandelbrot, "The Fractal Geometry of
Nature", Freeman, San Francisco (1982). J. Samuel, private
communications 1995. S. Srinivasan, A.K. Datye, M.H. Smith, and
C.'H.F. Peden, "Interaction of titanium isopropoxide with surface
hydroxyls on silica", J. Catalysis 145, 565-573 (1994).
10 A. Matsuda, Y. Matsuno, S. Katayama, T. Tsuno, "Weathering
resistance of glass plates coated with sol-gel derived 9TiO2.91SiO2
films", J. Mater. Sci. Lett., 8, (1989), 902- 904.
11 R. Sehgal, "Preparation and Characterization of Ultrathin
Sol-gel Derived Silica Membranes for Gas Separation Applications,"
M.S. Thesis, The University of New Mexico, 1993.
9